Problem. Let $G$ be a residually finite group, and identify $G$ with its image under the canonical map to its profinite completion $\hat{G}$. Let $x,y \in G$. Prove that the following conditions are equivalents:

(i) $x,y$ are conjugate in $\hat{G}$.

(ii) the imagens of $x,y$ in $G/K$ are conjugate in $G/K$, for every normal subgroup $K$ of finite index in $G$.

I'm sure that I'm confusing about this problem, so I'd like some help.

We know that $$\hat{G} = \varprojlim_{K \in I}G/K$$ where $I$ is a non-empty filter base of normal subgroups of finite index in $G$ and there is a continuous homomorphism $\varphi: G \to \hat{G}$ given by $\varphi(g) = Kg$. The pair $(\hat{G},\varphi)$ has the property:

"if $\psi: G \to H$ is a continuous homomorphism, to a finite group $H$, then there is a unique homomorphism $\overline{\psi}: \hat{G} \to H$ such that $\psi = \overline{\psi}\circ\varphi$."

  1. If $x,y \in G$, then $Kx,Ky \in \hat{G}$. So, the item (i) must be "... in $G$"?

  2. Who are the imagens of $x,y$ in $G/K$? Since $\hat{G} = \varprojlim G/K$, then $(G/K_{i},f_{ij})$ is a inverse system with inverse limit $(\hat{G},f_{i})$ where $f_{i}: \hat{G} \to G/K_{i}$? We can connect $G$ to $\hat{G}$ by $\varphi$ and $\hat{G}$ to $G/K_{i}$ by $f_{i}$, but who is $f_{i}$? I don't know a definition of completion using, explicitly, inverse systems indexed by a direct set and its maps. Thus, I don't know how to start this problem.

  • 1
    $\begingroup$ $\hat{G}$ is the set of sequences $(a_K K)_{K \in I}$ (indexed by finite index normal subgroups $K$) such that $a_K \in G, a_K K\in G/K$ and $a_K H = a_H H$ whenever $K\subset H$. Trying with $G = \Bbb{Z}$ and $I$ the index $p^n$ subgroups yields the $p$-adic integers, with $I$ every subgroup it is the profinite integers. $\endgroup$ – reuns Apr 21 at 20:28
  • 1
    $\begingroup$ Just a word on your profile: if you're tackling problems like this, then, surely, you'll be able to help someone. Have confidence. You're doing well :) $\endgroup$ – Shaun Apr 21 at 20:51
  • 1
    $\begingroup$ @Shaun thank you for the encouragement! $\endgroup$ – Lucas Corrêa Apr 21 at 21:04
  • 1
    $\begingroup$ You can also think to $\hat{G}$ as the set of (limits of) sequences $(g_n)_{n \ge 1}$ such that $g_n K$ ends being constant for every $K$, then $a,b\in G$ are conjugate in $\hat{G}$ if $a^{-1} (g_n b g_n^{-1}) K \to K$ for every $K$. $\endgroup$ – reuns Apr 21 at 21:31
  • 1
    $\begingroup$ $p$-adic integers $\Bbb{Z}_p$ are the completion of $\Bbb{Z}$ for the $p$-adic metric $d(b,b+a) = |a|_p = p^{-k}$ if $a \equiv 0 \bmod p^k,a \not \equiv 0 \bmod p^{k+1}$. Do you see how an equivalent metric can be defined in any group $G$ to obtain $\hat{G}$ as the completion ? $\endgroup$ – reuns Apr 21 at 21:50

You don't need a "filter basis of normal subgroups" to define $\hat{G}$. Take $\hat{G}= \underset{\leftarrow}{\mathrm{lim}} \: G/K$ over every normal subgroup of finite index $K$ of $G$. Then $\hat{G}$ is a subgroup of the product $\prod_K G/K$ (over the same $K$'s) as in the comment by @reuns, and one has projections $p_K :\hat{G} \to G/K$ for every normal subgroup of finite index $K$ of $G$, that are factorizations of the common projections $\pi_K : G \to G/K$ : $\pi_K = G \overset{i}{\to} \hat{G} \overset{p_K}{\to} G/K$ (where $i$ is the natural inclusion).

So $(i) \Rightarrow (ii)$ is pretty clear.

For the other inclusion, you have to build an element $z=(z_K)_K \in \hat{G}$ such that $zxz^{-1}=y$ from $z_K$'s such that $z_K p_K(x) z_K^{-1}=p_K(y)$ for every normal subgroup of finite index $K$ of $G$.

  • $\begingroup$ I think I got your idea! I will try. $\endgroup$ – Lucas Corrêa Apr 21 at 21:04
  • $\begingroup$ Actually you can keep the definition with a basis, you'll just need to recover any normal subgroup of finite index from the basis. It works the same way $\endgroup$ – AlexL Apr 21 at 21:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.